Speedup startup by caching last message timestamp

The cache is only approximate, i.e. it doesn't skip edits and similar,
but this might be good enough? Also no migration right now.

Speeds up startup by about 5x on my system.

Half the startup time is now loading the powerlevels for each room. We
can probably lazily load those too in the future.
This commit is contained in:
Nicolas Werner 2022-06-15 02:13:17 +02:00
parent 6e1fec1e63
commit ef9ebe3fd3
No known key found for this signature in database
GPG key ID: C8D75E610773F2D9
5 changed files with 95 additions and 23 deletions

View file

@ -353,14 +353,13 @@ TimelineModel::TimelineModel(TimelineViewManager *manager, QString room_id, QObj
, manager_(manager)
, permissions_{room_id_}
{
lastMessage_.timestamp = 0;
this->isEncrypted_ = cache::isRoomEncrypted(room_id_.toStdString());
auto roomInfo = cache::singleRoomInfo(room_id_.toStdString());
this->isSpace_ = roomInfo.is_space;
this->notification_count = roomInfo.notification_count;
this->highlight_count = roomInfo.highlight_count;
lastMessage_.timestamp = roomInfo.approximate_last_modification_ts;
// this connection will simplify adding the plainRoomNameChanged() signal everywhere that it
// needs to be
@ -1025,10 +1024,21 @@ isYourJoin(const mtx::events::Event<T> &)
return false;
}
DescInfo
TimelineModel::lastMessage() const
{
if (lastMessage_.event_id.isEmpty())
QTimer::singleShot(0, this, &TimelineModel::updateLastMessage);
return lastMessage_;
}
void
TimelineModel::updateLastMessage()
{
for (auto it = events.size() - 1; it >= 0; --it) {
// only try to generate a preview for the last 1000 messages
auto end = std::max(events.size() - 1001, 0);
for (auto it = events.size() - 1; it >= end; --it) {
auto event = events.get(it, decryptDescription);
if (!event)
continue;